Originally posted by: ArmenG

Hi,

Is CPLEX able to find the global optimum of any MIP problem? (or it is possible that CPLEX finds a local optimum of a MIP problem rather than its global optimum?)

Originally posted by: T_O

Up to numerical issues and defined tolerances, CPLEX will find the optimal solution for mixed integer linear programs (or get out of time or memory).

Originally posted by: ArmenG

My problem is an optimal scheduling problem. Due to the nature of the problem, as the scheduling horizon increases, the optimal value should not get worsen (i.e., either it should stay unchanged or it should improve). However, that is not the case for my problem. That's why I suspected that CPLEX is not able to find the global optimum and is stuck in a local optimum.

Based on your response, the only thing that I suspect is numerical issues. Is there a reference explaining about numerical issues resulting in non-optimal solution?

Originally posted by: T_O

This could also be a modelling error or an error in your assumptions.

Concerning numerical issues, you should have a look at kappa statistics.